EXTRACTION OF POSITIVE AND NEGATIVE ASSOCIATION RULES FROM TEXT: A TEMPORAL APPROACH

The focus of the association rule mining community, to a major extent, has been towards mining positive association rules. The negative association rules, counterparts of the positive association rules, have attained much less attention of the researchers. Finding the association rules from temporal perspective is quite new to the data mining community. The datasets used for mining temporal associations have generally been either data streams or market basket transactions. Extraction of temporal associations has not been too popular with the association rule mining researchers, especially the negative associations in temporal domain. The temporal association rules enable answering vital questions for the application of association rules, thus enhancing the trustfulness of the generated rules. Associations among diseases and symptoms, both positive and negative, are time/season dependent e.g. flu can have different symptoms in summer and in winter. The rule should positively represent some stable and reliable relationships in the domain. We propose a technique for mining temporal positive and negative associations from medical blogs. Experiments prove the efficacy of the proposed technique.